Abstract: In order to solve the target grouping below multi-frame data, this paper introduces a new grouping algorithm, TG-Stream, which can be divided into two parts: on-line part and off-line part. In on-line part, it uses the concepts of a pyramidal time frame and a Temporary Storage Structure(TSS) to save summary information of sensor data. In off-line part, it uses CNM algorithm to cluster the suitable data and output the grouping result. Experimental results show that TG-Stream algorithm has good equilibrium between accuracy and efficiency.

Key words: target grouping, multi-frame data, data stream clustering, situation assessment

摘要： 针对在多帧数据条件下的目标分群问题，提出一种基于数据流聚类的动态目标分群算法TG-Stream，该算法由在线和离线2个部分组成。在线部分采用临时存储结构(TSS)和金字塔时间框架保存侦察数据集的概要信息，离线部分采用CNM算法对时间框架的信息进行聚类，最终得到分群的结果。实验结果表明，TG-Stream具有灵活的精度和效率平衡性，能较好地满足决策辅助系统处理实时信息的需要。

关键词: 目标分群, 多帧数据, 数据流聚类, 态势估计

CLC Number: 

• TP18

• TP311